|
|
| GeoCommunity Mailing List |
| |
| Mailing List Archives |
| Subject: | RE: GISList: Geocoder comparisons * Build your own |
| Date: |
02/06/2003 07:55:52 PM |
| From: |
Neil Havermale |
|
|
There is no better speed-up to geocoding than clean data. If you want simple
geo-coding its generally included in the core product - no charge! The cost
is the accurate street information with accurate addressing as well as CLEAN
data that will match-up. Data cleaning prior to geoprocessing is a big
value added, aka "hits" and efficiency.
Andrew offers a smart and low cost method FYI.....
MidNight Mapper
Aka neil
-----Original Message-----
From: Canfield, Andrew [mailto:Andrew_Canfield@cable.comcast.com]
Sent: Thursday, February 06, 2003 12:28 AM
To: 'Alex Eshed': MapInfo-L (E-mail)
Subject: RE: MI-L Geocoding Scrubber!!!
I deal with this issue every day. Unfortunetly my solution will only work in
the US and Canada. There is an AI company called SemaphoreCorp that makes a
product called ZP4 it does everything from common spelling mistakes to
phoenetic string building. A lot of great programming went into the
Semaphores within the program.
The user interface side of it leaves a lot to be desired because it wasn't
really meant to be used as a stand alone product. You can do it that way but
it works best as a plugin. The great news for MapBasic people is that ZP4's
native connection language is DDE not OLE so you can plug it directly into
your MapBasic App. ZP4 has the USPS complete database compressed within
about four or five compression database files. It scrubs, then corrects and
gives you a zip + 4 hence the name. It will also do DPV validation based on
the postal records (DPV means delivery point validation, wether or not the
address has ever recieved mail before ). You can then run them through
MapMarker to do your geocode. What I did was canned the whole thing so my
users click a button browse to the text file of addresses then click run and
they end up with a table of scrubbed and geocoded addresses with no further
input. You have to also use the MapMarker API to do that.
The bare bones easiest way if you have to have it right now is scrub the
text file with ZP4, make a table of it with MapInfo send that through
MapMarker and your done in three steps rather than one but there is no
programming involved for those who don't want to or can't do the
programming. ZP4 is also really cheap and very fast I clean a million bad
addresses in ten to fifteen minutes on my machine. Longer if you are running
a slower machine but still very fast. I hate sounding like an add but I know
how frustrating it can be trying to find good cleaning software that doesn't
cost 30K so I will include the link to SemaphorCorp's site
http://www.semaphorecorp.com/cgi/zp4.html.
The Postal Database updates come once every two months if I remember
corrrectly so it runs about $950 a year but that's total cost there is no
startup or app fee the app is on the first cd every time they send it to you
if you need to reinstall it or whatever. It's a bit more expensive if you
run it on a network allowing everyone access to it but still not bad. The
next best one I found with that much AI in it was 30K to start and 6K a year
to maintain. With ZP4 it's always just 950 if you don't want DPV it's only
475 a year. I hope this helps. For a while over on Jaques's site he had an
example of my first app which canned the whole process. It is meant only as
an example of how one might choose to do this. So you can use it as source
code but it's not meant for production, only example. His site address is
http://www.paris-pc-gis.com/index.html and on the left frame about three
quarters of the way down is a link called files from other origins click
that and the center frame should change about halfway down that page is a
file called geomapbatch and it is meant as source examples only. Hope this
helps.
-----Original Message-----
From: Alex Eshed [mailto:Alexe@opisoft.com]
Sent: Wednesday, February 05, 2003 12:07 AM
To: MapInfo-L (E-mail)
Subject: RE: MI-L Geocoding Scrubber!!!
Greetings, List.
I'm sure this topic has occupied the minds of everyone involved in geocoding
by address.
My doubts coincide with Bill's. A very high level of artificial intelligence
would be needed to identify the geocodable address behind the "telegraph
mode" or "colloquial description."
So it seems that the most efficient way would be to employ a table of
synonyms to be scanned before geocoding. This is far more complex than it
sounds. It will work only in your language, and needs to address (no pun
intended) the most frequent typos, blunders and shorthand encountered.
We employ this method in our PreGeocoder. Several other factors are also
considered. For instance, the program suggests synonyms to the user when
none are found in the scan. Accepted suggestions are added automatically so
that the program "learns" by frequent usage. Several soundex-like tests are
also incorporated. Despite the fact that many "gibberish"
|
|
Sponsored by:
For information
regarding
advertising rates
Click Here!
|
